le même processus pour tous - Université de Bourgogne

702 M. Lelièvre et al. / Food Quality and Preference 19 (2008) 697–703
4.1. Are trained and untrained assessors comparable?
To address this question, we compared trained and untrained
assessors descriptions. In the condition without list, we found that
the descriptions of the groups of beers were rather different for
both groups of assessors. This result does not replicate Cartier
et al. (2006) study which found that the descriptions of groups of
breakfast cereals were almost similar between trained and untrained
assessors. We observed that there were many more terms
quoted by untrained assessors (54 terms) than by trained assessors
(35 terms). But when selecting only terms with a geometric mean
above 20%, only three terms for untrained assessors and eight terms
for trained assessors were kept. This result reflects the lack of consensus
in both the choice of the terms and in perceived intensity,
especially for untrained assessors. The greater lack of agreement
among untrained assessors in comparison to the trained assessors
is not very surprising. Indeed, training involves the development
of a common lexicon with standard physical references allowing
an alignment and a standardization of the sensory concepts of the
panelists (Ishii & O’Mahony, 1990). The importance of training in
reaching a consensus is illustrated by the fact that seven out of
the eight terms of trained assessors were attributes belonging to
the profile list of attributes used for their training. For example, a
trained assessor described the three Leffe beers in this way: ‘‘very
sweet, very alcohol, medium hop, medium bitter,” whereas an untrained
assessor described these same beers with: ‘‘medium exotic
feel, medium spicy sensation, medium grapping taste (goût prenant).”
This difference between the descriptors used by trained
and untrained assessors can be explained by the training of trained
assessors which allows them to possess a specific and precise
vocabulary. Finally we found that trained and untrained assessors
used the four intensity words differently. Contrary to untrained
assessors who used the three expressions ‘‘a little,” ‘‘medium,”
and ‘‘very” in the same way, we observed that trained assessors
used ‘‘very” twice as often as ‘‘a little.” We also noticed that untrained
assessors used the word ‘‘not” frequently, while trained
assessors hardly used it. So it seems that trained assessors tend to
describe their groups of beers with distinctive characteristics (i.e.,
characteristics with a high intensity) whereas untrained assessors
do not use particular characteristics to describe their groups of
beers. These observations highlight the interest of using intensity
scores to quantify attributes. These quantitative words bring additional
information to the descriptions and we think that it is important
to impose their use to the assessors.
The comparison between trained and untrained assessors’
descriptions confirmed the conclusions of several authors that
trained assessors used more specific terms, especially terms
learned during training (Chollet & Valentin, 2001; Chollet et al.,
2005; Clapperton & Piggott, 1979). We expected this high specificity
of trained assessors’ vocabulary to lead to a better matching
performance than that of untrained assessors. Yet, contrary to previous
work (Gawel, 1997; Lawless, 1984; Solomon, 1990) we did
not find any difference in matching performance between the
two groups of assessors. Both trained and untrained assessors were
above chance level but their performance levels were not very high
(54.7% of correct matches for trained assessors and 50.9% for untrained
ones). The overall low performance of trained assessors,
however, might be due to the high inter-individual variability. Indeed,
while three trained assessors performed perfectly, two others
were below chance level. A plausible explanation for this
high variability is the difference in years of training of the panelists.
Indeed, the panelists with 100% of correct matches were
among the panelists who had the longest training. Yet correlation
coefficient computed between the percentage of correct matches
and the years of training shows that it is not the only explanation
(r = .61, r 2 = 0.37, p < .05). The fact that some trained assessors with
four or five years of training succeeded in the matching task
whereas others had poor results may suggest that some trained
assessors are better than others to generalize their knowledge to
a new task. It has been already showed that trained assessors were
not able to generalize their perceptual knowledge to new beers
(Chollet et al., 2005). The same problem could exist with new tasks
and this might be related to the duration of training.
4.2. Is providing a list helpful?
We found that the descriptions of the beers were different
when assessors had a list of terms and when they did not have
such a list, especially for untrained assessors. For untrained assessors,
we observed a larger number of descriptors with a geometric
mean above 20% with the list than without the list. This suggests
that having a list of terms can be helpful for untrained assessors.
But a deeper look at the descriptions with the list shows that,
for example, untrained assessors used hop and malt to describe almost
all the beers. It is probable that the list given to untrained
assessors influenced their descriptions. The untrained assessors
probably knew that hop and malt are terms associated with the
brewing process and so they used it but without knowing exactly
what these terms mean. We assume that the descriptions containing
these words hop and malt did not allow them correct matches.
For trained assessors, the number of descriptors with a geometric
mean above 20% was quite similar between the two conditions.
Moreover, the results of the matching task were not better with
the list than without the list for both trained and untrained
assessors.
The efficiency of the list in this study can be put in perspective
with the results of Hughson and Boakes (2002). In this
study, assessors had to describe five white wines according to
three conditions: without any list of terms, with a long list of
terms (125 terms) and with five short lists of terms (14 terms
in each list corresponding to each wine). Then, they had to
match their own descriptions to the wines. Matching performance
was better in the short-list condition (40% of correct
matches) than in the long-list condition (27% of correct matches)
and in the control condition without any list (16% of correct
matches). Moreover, only results in the short-list condition were
above chance. So we can wonder why our list did not help assessors
to improve their scores of matching too. One reason could
be that our list of terms was too long (44 terms) compared to
the one of Hughson and Boakes (14 terms) to help assessors to
effectively describe the beers. In the case of trained assessors,
another reason could be that the terms provided were different
from the terms used in training. This hypothesis is supported
by the fact that trained assessors described ChtiBL as butter in
the condition without the list but did not in the condition with
the list. Interestingly, in Meilgaard’s list, butter is replaced by
diacetyl, which is associated with the butter flavor and so trained
assessors did not seem to know the term diacetyl. This remark
highlights the importance of using a common descriptive vocabulary.
Some authors such as Rainey (1986), Civille and Lawless
(1986) or Stampanoni (1994) indicated that for sensory profiles,
the use of a common terminology based on references reduced
the time for training and improved the agreement between the
assessors. In our case, the use of a terminology without associated
reference did not help assessors to describe the beers.
Finally, the fact that the list of terms did not help the assessors
could be due to the use of a previously published list which was
not exactly adapted to our products. In the study of Hughson
and Boakes (2002), the short lists provided to the assessors contained
terms which corresponded exactly to the wines to be
described.